Assessment for Production and Operation Ability of Medium and Small-sized Enterprises Based on Neural Network

In order to improve production and operation ability of medium and small-sized enterprises, an assessment-index system of production and operation ability was proposed, and a corresponding assessment model was established based on BP neural network. The conjunction weights of the neural network were continuously modified from output layer to input layer in the process of neural network training to reduce the errors between the anticipated and actual outputs. The results from an example show that this method is reliable and feasible. The production and operation abilitv of an enterorise with assessed result of 0.833 is fairly oowerful, and that with assessed result of 0.644 is average.

Recycling Strategy and Recyclability Assessment Model Based on the Artificial Neural Network

Now, a rapidly growing concern for the environmental protection and resource utilization has stimulated many new activities in the in dustrialized world for coping with urgent environmental problems created by the steadily increasing consumption of industrial products. Increasingly stringent r egulations and widely expressed public concern for the environment highlight the importance of disposing solid waste generated from industrial and consumable pr oducts. How to efficiently recycle and tackle this problem has been a very impo rtant issue over the world. Designing products for recyclability is driven by environmental and economic goals. To obtain good recyclability, two measures can be adopted. One is better recycling strategy and technology; the other is design for recycling (DFR). The recycling strategies of products generally inclu de: reuse, service, remanufacturing, recycling of production scraps during the p roduct usage, recycle (separation first) and disposal. Recyclability assessment is a very important content in DFR. This paper first discusses the content of D FR and strategies and types related to products recyclability, and points out th at easy or difficult recyclability depends on the design phase. Then method and procedure of recyclability assessment based on ANN is explored in detail. The pr ocess consists of selection of the ANN input and output parameters, control of t he sample quality and construction and training of the neural network. At la st, the case study shows this method is simple and operative.

A Condition States Assessment System for Concrete Bridges Using Neural Networks

Due to continuing aging and heavy utilization of many bridges and the limited available funds, the importance of proper bridge condition state assessment has risen recently, which is the crucial point for rational decision-making on MR&R activities. This paper presents a prototype of the concrete bridge condition state assessment system (CBCSAS) with the following sub-modules: inspection, parameter recognition, structural assessment, main cause identification and priority-to-action. And multi-layer neural networks, which may combine with fuzzy set theory or not, are performed to realize the structural assessment with embedding expert knowledge into the assessment system.

Fuzzy neural and chaotic searching hybrid algorithm and its application in electric customers’s credit risk evaluation

To evaluate the credit risk of customers in power market precisely, the new chaotic searching and fuzzy neural network (FNN) hybrid algorithm were proposed. By combining with the chaotic searching, the learning ability of the FNN was markedly enhanced. Customers’ actual credit flaw data of power supply enterprises were collected to carry on the real evaluation, which can be treated as example for the model. The result shows that the proposed method surpasses the traditional statistical models in regard to the precision of forecasting and has a practical value. Compared with the results of ordinary FNN and ANN, the precision of the proposed algorithm can be enhanced by 2.2% and 4.5%, respectively.

Damage Assessment of Existing Transmission Structures Using ANFIS （Adaptive Neuro-Fuzzy Inference） Model

This paper introduces a new methodology for the damage assessment of existing-transmission structures using six layers, zero order Sugeno model. The model is a hybrid fuzzy-neural system that combines the power of neural networks and fuzzy systems. It is a learning expert system that finds the parameters of the fuzzy sets and fuzzy rules by exploiting approximation techniques from neural networks. The condition ratings of the structural components are determined based on visually observed deterioration-symptoms and the severity of those symptoms. A supervised learning process using training data and expert opinions is used to develop the expert system rules and determine the ratings of the structural components. For the learning from training data, the model uses a combination of least-square estimator and gradient descent method. A sequential least square algorithm is used to determine the weighting factors that minimized the errors. A test case is given to illustrate the power of the proposed fuzzy-neural system. It is concluded that the Sugeno model＇s ability to tune the parameters based on the training data makes it superior to the rules produced by an expert in the conventional fuzzy logic systems.

基于统计和自适应ParNet的产学研绩效评价

针对现有产学研绩效评价体系及方法中存在的评价指标覆盖范围单一、评价样本特征表达不充分、评价模型自优化能力待提高的问题,提出主客观产学研综合绩效智能评价的评价体系及方法。首先,围绕三方合作主体,挖掘产学研合作过程中影响绩效的要素及这些要素之间的联系,自主构建主客观产学研绩效三级评价体系;其次,通过将收集到的离散序列评价样本映射至极坐标空间、马尔可夫转移矩阵等不同高维空间域,增强离散样本特征表征;然后,通过基于精英反向翻筋斗觅食的混沌优化策略设计,提高深度模型冗余压缩和超参数的全局寻优效率,构建轻量压缩及高维超参数的自适应寻优的ParNet(AParNet)分类模型;最后,将模型应用于产学研绩效评价中,实现高性能的绩效智能评价。实验结果表明,所提方法很好地贴合了离散序列非线性分类应用,同时模型中加入优化策略后,在减少计算量的同时提高了分类性能,具体体现在:与ParNet相比,AParNet中的参数量减少了10.8%,较好地实现了模型的压缩,且它在产学研绩效评价中的分类准确率可达到98.6%。在产学研绩效智能评价应用中,该方法提高了评价模型的自适应能力,能够实现准确、高效的产学研绩效评价。

应用于水下防御的小目标动态多属性威胁评估

面向近海水下防御的信息融合,提出一种危险小目标的动态多属性威胁评估方法。通过对典型目标的特性分析,构建了威胁评估指标体系。利用K均值聚类为各规范化指标添加威胁度标签,从而生成指标值-威胁度样本。基于此,将指标威胁隶属度函数的确定和计算转化为径向基函数神经网络的训练和预测,连接权值及偏置等网络参数利用量子粒子群优化算法求解,目标属性及时间序列的权重分别由熵权法和正态分布累积函数计算,应用逼近理想解排序法得到目标威胁度。算例分析中,通过专家评分和对比实验验证了所提方法的有效性和可靠性。

A novel cascaded H-bridge photovoltaic inverter with flexible arc suppression function

This paper presents a novel approach that simultaneously enables photovoltaic(PV)inversion and flexible arc suppression during single-phase grounding faults.Inverters compensate for ground currents through an arc-elimination function,while outputting a PV direct current(DC)power supply.This method effectively reduces the residual grounding current.To reduce the dependence of the arc-suppression performance on accurate compensation current-injection models,an adaptive fuzzy neural network imitating a sliding mode controller was designed.An online adaptive adjustment law for network parameters was developed,based on the Lyapunov stability theorem,to improve the robustness of the inverter to fault and connection locations.Furthermore,a new arc-suppression control exit strategy is proposed to allow a zerosequence voltage amplitude to quickly and smoothly track a target value by controlling the nonlinear decrease in current and reducing the regulation time.Simulation results showed that the proposed method can effectively achieve fast arc suppression and reduce the fault impact current in single-phase grounding faults.Compared to other methods,the proposed method can generate a lower residual grounding current and maintain good arc-suppression performance under different transition resistances and fault locations.

含分布式电源的新型配电网承载力模糊综合评估及提升策略研究

随着大量分布式电源广泛并入配电网,新型配电网的承载力评估成为系统安全运行的关键,因此有必要对含分布式电源的新型配电网承载力进行综合评估并在此基础上考虑适当的提升策略。首先,建立了新型配电网分布式电源承载力的评估模型,并采用二阶锥松弛技术加快模型的求解速度。其次,建立了多维度多层次的新型配电网承载力实时评估体系,根据得到的各指标动态综合权重,采用模糊综合评价法计算配电网承载力的综合得分。最后,提出了一种基于改善电压约束的配电网承载力提升策略。在IEEE 33节点系统配电网模型进行算例分析,通过比较所提方法和传统方法下配电网承载力变化情况及综合评估分数,证明了所提提升策略的有效性和评估体系的科学性。

基于模糊神经网络的一体化监测设备可靠性评价

为了实现对一体化监测设备的可靠性综合评价,基于T-S模糊神经网络的方法,以时效性、粗差占比、采样间隔、误报次数4个指标作为输入向量,利用T-S模糊系统的数据处理能力、神经网络的非线性拟合能力及MATLAB软件编程,构建一体化监测设备的综合评价模型。将训练好的T-S模糊神经网络模型与BP神经网络模型同时用于6组监测设备的预测,从数值模拟与应用的情况可以看出,普适型设备略强于一体化监测设备,与实际情况相符。该方法在一体化监测设备的评价方面具有可行性和实用性,可以作为该设备在实践应用过程中可靠性的依据。

基于改进模糊神经网络的细胞活性预测

针对细胞活性的预测问题,提出了一种利用自适应差分进化算法优化的模糊神经网络(ADE-FNN)预测模型。首先,通过模糊化处理细胞培养过程中的多种影响因素(如培养基成分、温度、pH值等),将这些因素作为模糊神经网络的输入。然后,利用已有的细胞活性数据对模型进行训练,优化网络参数。经过多次迭代和调整,模型逐渐学习到输入与输出之间的复杂映射关系。最后,利用细胞活性仿真实验验证所提ADE-FNN算法的性能。结果表明,基于模糊神经网络的细胞活性预测模型具有较高的预测精度和泛化能力。与传统的统计方法相比,该模型能够更好地处理数据中的不确定性和噪声,从而提供更准确的预测结果。此外,该模型还具有较好的可解释性,有助于深入理解细胞活性的影响因素及其作用机制。

Tremor suppression strategy of the tremor suppression robot

Based on the development of the performance of magnetorheological damper,the tremor suppression strategy that uses angle displacement sensor and force sensor to sense the information of the tremor and uses the fuzzy neural network as the control algorithm is presented. The information of the maximum tremor torque can be achieved exactly by means fusing the data of displacement sensor and the data of force sensor. Due to the slow change of the maximum tremor torque,the fuzzy neural network with single input variable and single output variable is adopted as the control algorithm,and the mathematic model of fuzzy neural network is worked out. The elbow tremor suppression model is established to provide tremor information for fuzzy neural network training and tremor suppression strategy simulation,then the off-line fuzzy neural network training is finished with the selected training data set,and the tremor suppression strategy is simulated. Simulation results show that the tremor suppression strategy has presented is effectually for the tremor suppression.

陶瓷浆料3D打印机挤压力模糊神经网络PID稳定控制研究

针对在微流挤出陶瓷浆料3D打印机作业过程中挤压力稳定控制的需求,根据打印机挤压力控制系统非线性、时变性的特点,总结了现有挤压力稳定控制策略的优缺点,并在模糊PID (proportion-integral-derivative,比例-积分-微分)控制器中嵌入神经网络结构,提出了挤压力模糊神经网络PID稳定控制策略。该策略基于六层模糊神经网络,以挤压力偏差值e和偏差值变化率ec为输入,PID控制器控制参数为输出,完成正向模糊控制过程,并基于神经网络的自学习优势实现反向传播及在线更新神经网络权值,以实现打印过程中挤压力的精准自适应调节。挤压力控制Simulink仿真、挤压力控制实验及坯体打印实验表明:相较于传统PID控制策略,采用模糊神经网络PID控制策略可使超调量减小20.9%,挤压力提前90 s达到稳定状态,压力峰值减小12 N,压力谷值增大18 N;相较于采用模糊PID控制策略,超调量减小1.73%,挤压力提前56 s达到稳定状态,压力峰值减小4 N,压力谷值增大8 N;模糊神经网络PID控制策略具有一定的优越性,可使打印过程中挤压力的控制精度更高,稳定速度更快,超调量更小,所打印坯体的整体形貌质量更优,也可使控制系统的鲁棒性更好。研究结果为其他工业设备的PID控制、智能控制提供了新的思路和方法。

基于轻量级卷积神经网络的科技型中小企业估值研究

对科技型中小企业进行科学合理估值有利于其募集资金从而健康发展,但传统估值方法的准确度和适用性均有待考量。鉴于此,利用文献分析与关键词提取方法,初步从相关文献中提取科技型中小企业估值指标,并通过德尔菲法筛选得到能体现科技型中小企业核心能力的关键指标;进一步设计轻量级卷积神经网络(TecNet)结构,基于大数据训练构建适用于科技型中小企业的估值模型;最后以10家样本企业为例,对其2017—2021年度预测值与真实值相对误差进行拟合分析以及敏感度分析,以验证所构建估值模型的有效性。结果表明:模型预测的拟合曲线与真值重合度较高,平均相对误差为3.43%,且90%的样本的相对误差在10%以内,说明Tec Net模型具有科学有效性;将关键指标提升5%时,硕士及以上比例、研发投入占营业收入比例以及专利数对估值结果浮动程度最高,分别为16.51%、16.10%和11.87%,表明科技型中小企业价值对这3项指标最为敏感。因此,科技型中小企业发展应着眼于其关键指标,通过优化员工学历层次、加大研发投入等措施提高智力资本,并持续加强技术创新。

基于反向传播神经网络的海洋工程项目投标风险评价方法

海洋工程行业是一个国际化的行业,其国际化的性质决定了该行业的竞争的激烈性.同时海洋工程项目需要大规模的资金投入,所以在进行该类项目投标时,公司决策人员能否对拟投标项目进行正确的风险评估后做出合理的投标决策,对公司的长期发展起到至关重要的作用.通过对海洋工程行业投标项目相关风险因素的分析和识别,使用模糊层次分析法量化定性问题,并通过反向传播(Back Propagation,BP)神经网络的高容错、高泛化能力,建立相应项目风险评价模型.经过验证,该模型评估结果与实际专家评分结果相一致,具有较高的准确性,从而为海洋工程行业的配套企业在投标决策过程中提供了一个有效且快速的风险分析工具.

基于数据挖掘技术的企业审计模型构建及算法优化

采用层次分析法,从重大报错风险、审计人员风险、审计程序风险3个维度构建了电力企业审计风险评估指标,对BP神经网络初始权值、阈值进行优化,提出了电力企业审计风险评估的BSA-BP模型。将提出的模型和专家评估结果进行对比,2种方法保持了一致性。采用BAS-BP神经网络找出了电力企业审计工作开展过程中的薄弱环节,采取针对性的措施进行改进。2次审计评估结果表明,采取针对性的措施改进后,企业的审计风险明显降低。

基于MSADE-IT2FNN模型的软测量建模方法及应用

针对复杂工业过程某些关键参数无法有效、实时在线检测的问题,提出一种基于多策略、自适应差分进化算法(MSADE)优化的区间二型模糊神经网络(IT2FNN)软测量建模方法。首先,为了解决差分进化算法采用单一策略、固定缩放因子和交叉概率导致后期搜索能力不足的问题,提出一种多策略、自适应的差分进化算法(MSADE),该算法利用IT2FNN模型的均方根误差(RMSE)作为适应度函数,通过搜索不同规则数下的RMSE值,从而确定IT2FNN的结构(规则数)和初始参数;然后,IT2FNN模型的参数利用梯度下降法进行学习。最后,将所提模型应用到Mackey-Glass混沌时间序列的预测和酿酒过程淀粉利用率的软测量建模问题中,仿真结果验证了提出方法的有效性和优越性。

电力企业信用度综合智能评分及风险等级评价

针对电力企业信用度性能差,风险评估能力差等问题,提出了一种新型的管理方法,构建了一种新型的综合智能评分架构方案。该方案能够融合多种数据信息,实现电力企业信用度综合智能评分。在该电力企业信用度综合智能评分系统的基础上,设计了改进型模糊评价方法,该方法融合了DA-BP神经网络模型,又通过卷积分计算,提高了电力企业信用度评估能力。通过试验,本研究的方法评估精度高,性能稳定可靠。

基于驾驶员驾驶意图的混合动力汽车能量管理策略研究

驾驶员的驾驶意图和风格识别是确定汽车(特别是混合动力汽车)控制策略的理论基础。基于高精度驾驶意图识别的能量管理策略可以合理切换发动机和电机的使用时机,从而在提升驾驶员操作体验的同时降低车辆百千米油耗和排放量。使用实测道路数据构建了驾驶意图识别模型,基于Cruise搭建了单轴并联式混合动力汽车模型,并根据模糊控制理论基于Simulink构建了模糊控制策略。在NEDC工况下分别应用三种驾驶风格控制策略和综合控制策略进行联合仿真,验证了模糊控制策略能够将SOC值控制在合适的区间,提高汽车的经济性、排放性。

基于BN-FCEM的化工企业安全风险分析方法

近年来,随着中国化工产业的迅速发展,化工企业危险品化学灾害事故仍时有发生。为更好地提高化工企业安全生产水平、精准识别化工企业中的重要风险因素并合理规避风险,提出一种基于贝叶斯网络(BN)和模糊综合评价法(FCEM)的化工企业安全风险分析方法。首先,该方法基于人、物、技、管、环五个方面共20个与化工企业相关的风险因素,建立了化工企业安全风险BN模型;然后,以模糊理论集(FTS)表征BN模型中概率参数的专家评估信息,结合Netica软件分析基于BN的化工企业安全风险概率和重要风险因素,并采用FCEM评估化工企业的整体安全风险损失;最后,综合考虑化工企业风险概率和风险损失,得出化工企业的安全风险等级。为验证该方法的可行性,以舟山某化工企业为例,结合风险等级矩阵,评估了该化工企业的安全风险,并提出了相应的改进建议。